How Did Trump Win?

An extraordinary new paper by Justin Grimmer and William Marble at Stanford has totally and irretrievably debunked the racial resentment thesis that traced the catastrophe of 2016 to white racial prejudice. But the paper, “Who Put Trump in the White House? Explaining the Contribution of Voting Blocs to Trump’s Victory,” does much more than that. It explains why the vast bulk of the literature that has emerged got it so very wrong. And it does so by mathematically demonstrating the limitations and biases of previous analyses in a straightforward manner that is a model of simplicity and elegance. This is easily the most significant work to appear on the question. In many ways, it is as much a theoretical intervention as an argument over 2016; one that has all the hallmarks of a seminal work ā€” that creates a before and after. And it has the potential to irrevocably change the conversation in both academic political science and sophisticated political consulting. So what have Grimmer and Marble shown?

They begin by noting that, in order to understand 2016, or any other election, it is not enough to show that voters with such and such attribute (denoted by x, eg racial resentment) voted for this candidate at higher rates. This is so because the effect may be swamped by compositional effects (ie, the share of people with that attitude in the population may have fallen) and turnout rates (ie, the people with that attitude may have turned out at lower rates). In order to understand how a candidate won, we must pay careful attention to all three factors at once: composition, turnout, and vote choice.

The number of votes that Trump received from voting bloc (ie, whatever attribute) x is given by the product of (1) the share of the electorate in voting bloc x, (2) the turnout rate conditional on voting bloc x, and (3) the rate at which they voted for Trump conditional on turnout and bloc. This a mathematical fact, there is no arguing with it:


The problem with the vast bulk of the literature is that it pays no attention to these confounding effects and pays near-exclusive attention to the vote choice of various blocs (“authoritarians” &c). In a survey of 83 papers analyzing 2016, they found a mere 5 that had paid attention to all three. The vast majority of reported results, 94 percent, are suspect because they fail to take into account these mathematical facts. This includes the entirety of the vast literature supporting the racial resentment thesis.


Once you start adding up the correct way, the racial resentment thesis turns out to be flat out wrong. The paper contains a number of figures of the sort that look like the following. Here’s how to read them quickly. Please read this carefully so that you can interpret the graphs that follow.

Start at the bottom-right panel, (f) Composition, that shows the proportion of the electorate in the buckets and how that changed between 2012 and 2016. In this figure the buckets are quintiles of racial resentment. We can see that the proportion of the electorate in the highest quintile of racial resentment fell significantly, while that in the lowest two quintiles increased. Next, look at the lower-middle panel, (e) Turnout, to see the turnout rates. We can see that turnout was lower for the most racially resentful. Then see the bottom-left panel, (d) Vote Choice, that reports the probability of swing to Trump for the bloc conditional on turnout. Finally, look at the top-left panel, (a) Change in Relative Support, that combines these three factors into a net effect on the Trump swing (ie Trump less Romney). We can see that the swing to Trump cannot be attributed to racial resentment because the net effect was higher for middle quintiles by racial resentment than the highest quintile. The other two panels, (b) Net Republican Votes and (c) Number of votes, sum up the aggregate net votes by bloc for Trump and Romney, displayed in (b), and the same as a share of the voting eligible population (VEP), displayed in (c). The bottom line is always at the top-left, the Trump swing by attribute computed correctly.


A different proxy for “deplorables” is attitudes towards immigration. There again, we find that the big swings were in the second and third quintile of opposition to immigration. This was so even though the electorate shifted against immigration in 2012-2016.


The results are robust to alternate specifications. For instance, using continuous attributes instead of discrete quintiles.

Immigration smooth.png

Yet another proxy for “deplorables” is attitudes towards gay marriage. Here again, we find that the Trump swing can be traced to supporters of gay marriage not those opposed.


A still another proxy for “deplorables” is attitudes towards Muslims. Again, no surprise here, people who feel resentful towards Muslims did not vote for Trump in higher numbers than Romney. To the contrary. muslims smooth.png

They obtain similar results for swing states and for voters who voted for Obama in 2012. Among Obama votes, they find the highest swing among the second lowest quintile by opposition to immigration.


Now the racial resentment thesis has been thoroughly debunked, what actually accounts for the Trump swing? Turns out, the Policy Tensor had hit the nail on the head: The bloc who put Trump into the White House was the white working class. Trump is an expression of the breakdown of elite-mass relations in the United States. The strongest swing for Trump was among those with only a high school diploma, and the weakest was among those with a college degree ā€” recall that these are the structural parameters of American class society.


Similar results obtain if we use income quintiles. The Trump swing was the greatest in the bottom three quintiles, which as we have seen, is where the white working class resides.

income close.png

The primacy of the class axis is equally evident in the swing states, defined here as those states where the margin of victory was less than 5 percent. The results are identical to that for the country as a whole. This is not something that happened only in the swing states. The revolt of the white working class, of which Trump is merely a symptom, was across the board, and in no way confined to the postindustrial swing states that have borne the brunt of the decline of US manufacturing employment.

class close.png

It would be interesting to combine this analysis with the geography of the electoral college ā€” a crucial conditioner of 2016, if not every modern election. It would also help to take the next step and ask: Given that we now know who put Trump in the White House, why did they do so?

Gimmer and Marble do not address why the blocs voted as they did. We have offered an interpretation that is the only one consistent with facts they have established. Trump is in the White House because of the breakdown of elite-mass relations; which in turn is due to the rise of Boasian antiracism as the hegemonic ideology of the social elite. It is the clash of working class and elite ideologies that has driven the breakdown of elite-mass relations. Moreover, it is precisely the grip of Boasian antiracism on the minds of the social elite or the professional class that made the now debunked racial resentment thesis so compelling to the professoriate and the scribes.

Given that a restoration of elite-mass relations is a prerequisite for a progressive-working class alliance, and the latter is a prerequisite for an exit from the the planetary impasse, the acceptance of this diagnosis, and a formulation of a path forward on that basis, is no longer optional.

Many thanks to Gimmer and Marble for giving me permission to share their graphs on these pages.


4 thoughts on “How Did Trump Win?

  1. The widespread revulsion toward elite ideology guarantees Trump the election, unless the elites start WWIII and impeach him first as unfit for wartime command. See

    However, even if he wins, the big money still owns Congress, and progressive tax reform is unlikely. Even Bill Gates is saying rich people should pay more in taxes, and is open to a wealth tax! But what Congressperson is going to vote against his or her donors?

    Interesting times.

  2. The alleged indicators of “racial resentment” do not actually measure anything to do with racism. Not any tendency to discriminate in action:

    Click to access 40_paper.pdf

    Nor in thought. People (except US progressives/liberals “warm” to African-Americans) do not give different answers to the “racial resentment” questions, give the same answer no matter what group you ask the question about.

    Click to access carneyenos.pdf

    Fun bonus fact, warmness to your own group does not correlate with negativity to other groups.

Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s